Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package python-vllm for openSUSE:Factory checked in at 2026-09-10 11:49:38 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/python-vllm (Old) and /work/SRC/openSUSE:Factory/.python-vllm.new.1265 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-vllm" Thu Sep 10 11:49:38 2026 rev:9 rq:1376831 version:0.29.0 Changes: -------- --- /work/SRC/openSUSE:Factory/python-vllm/python-vllm.changes 2026-08-27 18:55:10.311530831 +0200 +++ /work/SRC/openSUSE:Factory/.python-vllm.new.1265/python-vllm.changes 2026-09-10 11:52:17.852871066 +0200 @@ -1,0 +2,30 @@ +Thu Sep 10 05:45:10 UTC 2026 - Martin Pluskal <[email protected]> + +- Update to version 0.29.0: + * Model Runner V2 is now the default runner for all models, with + CUDA graph memory profiling for KV cache auto-sizing and + batch-sharded sampling cutting per-step logits memory + * New models: Hy4-preview, Qwen3.8-Flash-Next, GraniteSWA/MoeSWA, + NemotronH_Omni_Reasoning_V3; Kimi K3 and DeepSeek V4 performance + push across kernels, MLA and speculative decoding + * Breaking changes: ten deprecated model architectures removed, + FlexOlmo/Olmo3/Hunyuan moved to the Transformers backend, PyAV + video decoder backend removed, api_server module deprecated in + favor of vllm serve + * Security hardening: cache_salt length bounded against scheduler + CPU exhaustion, oversized media rejected before download, audio + clip size enforced on all audio paths, api_key/hf_token redacted + from logs + * Many further model, serving and scaling changes -- see upstream's + release notes for the full list +- Follow the new upstream floors: python-transformers >= 5.10.4 and + python-huggingface-hub >= 1.28.0, both already exceeded by what the + devel project ships +- Re-test both downstream patches against 0.29.0, no rebase needed: + requirements carry the same exact pins, and setup.py keeps the + rust_extensions block the Rust-frontend patch keys off +- The oneDNN and Arm Compute Library pins in + cmake/cpu_extension.cmake are unchanged, so the vendored Source + tarballs carry over untouched + +------------------------------------------------------------------- Old: ---- vllm-0.28.0.tar.gz New: ---- vllm-0.29.0.tar.gz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ python-vllm.spec ++++++ --- /var/tmp/diff_new_pack.5Ibnef/_old 2026-09-10 11:52:20.280972904 +0200 +++ /var/tmp/diff_new_pack.5Ibnef/_new 2026-09-10 11:52:20.282972988 +0200 @@ -48,7 +48,7 @@ %define acl_version 52.6.0 %bcond_without libalternatives Name: python-vllm%{psuffix} -Version: 0.28.0 +Version: 0.29.0 Release: 0 Summary: A high-throughput and memory-efficient inference and serving engine for LLMs License: Apache-2.0 @@ -119,7 +119,7 @@ Requires: python-einops Requires: python-fastapi >= 0.133.0 Requires: python-filelock >= 3.16.1 -Requires: python-huggingface-hub >= 1.27.0 +Requires: python-huggingface-hub >= 1.28.0 Requires: python-ijson Requires: python-jsonschema >= 4.23.0 Requires: python-lark >= 1.2.2 @@ -172,7 +172,7 @@ # there and only makes the package uninstallable until it is bumped by hand. Requires: python-torch >= 2.12.0 Requires: python-tqdm -Requires: python-transformers >= 5.5.3 +Requires: python-transformers >= 5.10.4 Requires: python-typing_extensions >= 4.10 Requires: python-watchfiles Requires: python-xgrammar >= 0.2.1 @@ -289,7 +289,7 @@ rm -f $sd/vllm/vllm_flash_attn/.gitkeep # These modules carry a #!/usr/bin/env python shebang but are imported, not run. sed -i '1{/^#!/d}' $sd/vllm/entrypoints/grpc_server.py -sed -i '1{/^#!/d}' $sd/vllm/entrypoints/openai/dp_supervisor.py +sed -i '1{/^#!/d}' $sd/vllm/entrypoints/launchers/dp_supervisor.py %fdupes $sd } %python_clone -a %{buildroot}%{_bindir}/vllm ++++++ vllm-0.28.0.tar.gz -> vllm-0.29.0.tar.gz ++++++ /work/SRC/openSUSE:Factory/python-vllm/vllm-0.28.0.tar.gz /work/SRC/openSUSE:Factory/.python-vllm.new.1265/vllm-0.29.0.tar.gz differ: char 5, line 1
